การจำแนกชั้นเนื้อของมะพร้าวน้ำหอมอ่อน

• Main Authors: ธเนศ เคารพาพงศ์, ศุภชัย มะเดื่อ
• Other Authors: Faculty of Engineering Computer Engineering
• Format: Technical Report
• Language: Thai
• Published: มหาวิทยาลัยสงขลานครินทร์ 2022
• Subjects: มะพร้าว การเก็บเกี่ยว; เทคโนโลยีการเกษตร
• Online Access: http://kb.psu.ac.th/psukb/handle/2016/17376
• Institution: Prince of Songkhla University

This research proposes a method for young aromatic coconut classification from an image of the bottom part of the coconut using image processing techniques. Coconuts are classified into 3 categories: single layer, one-and-a-half layer and double layer. Preliminary experiments were conducted to search for significant color components in 3 color standards: RGB, HSV and CIELAB. Percentage of the color area inside interest region was used as a classification threshold. Experimented results showed that the plane in HSV color standard and the A plane in CIELAB color standard were significant. Their intensities are correlated with coconuts age. In classification phase, the minimum distance interested regions for classification are defined. Circular rings around the center of the coconuts for each ring the percentage of white pixels with are computed. Then the relation between the ring order and the percentage of the white pixels of each ring and plotted. The graph is approximated by a second-order polynomial function. Each coconut category has its own polynomial function. In classification mode, the graph of the unknown sample is compared against these 3 polynomial functions. The function that yield the minimum distance is the answer. From experiments, the image in S plane of the HSV color space yields the highest accuracy, that is 88.89% for single layer coconuts. 86.67% for one and-a-half layer coconuts and 85.95% for double layer coconuts. This yields the overall accuracy of 86.95%